Samba and Oracle Archived Redo Logs

[Don Spare]

I am running out of space to store archived redo logs on my L2000 box and am considering using Samba to make the archive destination on a NT box that has more than enough space to support this.
My question is: What are the repercussions of this? Will I be able to use these files to recover the database? Do they survive the translations to NT and back without problems?

If anyone has any experience with this I would appreciate it.

Don

[Tom Jackson]

Hi Don:

Redo log files record almost all changes made to data in the database buffer cache. They are used when an instance fails to recover commited data that has not been written to the data files. When determining the location of redo log files, Oracle recommends that you place different members of a group on different disks; no mention of networked disks. You only really need the current redo log file. The archived ones just audit the DB's transactions. You can move archived redo logs to another disk, or on the network, but I would keep the log groups and members locally.

Tom

[Steven E. Protter]

You are better off frequently moving unneeded files off the HP server.

Reason: performance.

Writing archive logs directly to NT goes through a NIC card which is slower than disk access in most cases.

You can probalby do what you want, but I'd research it on metalink.oracle.com

Attacched is a tested, production script that you will need to modify.

Steve

[Rick Garland]

Being able to extend the archlogs filesystem would definitely be the way to go.
If not possible look at compressing the archlogs on a more frequent basis.

I would avoid going to a different system and/or across the network. It will work but time can sometimes be of the essence and with this method you are saying time can wait.

[Raynald Boucher]

Yes you can use an NT (or Linux) box, just remember to use binary ftp.

Set up a cron job to run periodically (time span should equal creation of 2 archived logs max) that will first ftp and compress all uncompressed logs.

Set up a second cron job (daily) to purge all archived logs that are older than your security window (ex. 3, 4 or 5 backups ago).

Set up a third job to age the backed up logs on the destination server.

Placing the destination server off site will reduce your data loss in case of disaster recovery.

[W.C. Epperson]

Disclaimer: have not been Oracle certified since V6.

You DO need the archived logs in a scenario where you've physically lost the current database and have had to restore a previous copy from backup. If you have all the redo logs back to the date/time of the backup, you can roll forward to the point in time of the crash.

Archive logs are not written to continuously, but only when an online log switchover occurs, so you wouldn't see continuous performance problems. However, when an online log switchover happens, everything (DBMS-wise) stops until the switchover completes, meaning the archiving must complete first. When archiving to slow media or across a slow network, you'll have the appearance of periodic "hangs" while archiving takes place. And if your Samba server is down, the "hang" will last until it comes back up.


What I always did was have a weekly backup job that put ALL the current archived logs on tape, then purged the ones more than two weeks old.